1/ AI gets caught because its writing can be predictable.
Detectors measure how uniformly the text flows. Same sentence length, same shape, line after line. This uniformity is what indicates it.
GENERATIVE AI
-
AI betrayed by the uniformity of its writing
By
–
-
Rhythm betrays AI: how to fix it
By
–
Everyone can tell when you used AI to write. It's not the words. It's the rhythm. Here's how to fix it:
-
New benchmark MPMWorlds tests AI understanding of physics in videos
By
–
AI can now watch a video and write the physics code behind it.
— AlphaSignal (@AlphaSignalAI) 16 juin 2026
Video models are often called world simulators.
A new paper tests whether they actually understand physics.
MPMWorlds is a benchmark of 95,000 rendered 2D simulations.
It covers liquids, snow, sand, and… pic.twitter.com/b95y3sdWIAAI can now watch a video and write the physics code behind it. models are often called world simulators. A new paper tests whether they actually understand physics. MPMWorlds is a benchmark of 95,000 rendered 2D simulations. It covers liquids, snow, sand, and
-
3B Model VibeThinker-3B Nears State-of-the-Art on Reasoning
By
–
Crazy: A 3B model is now reaching highly competitive results on verifiable reasoning tasks. VibeThinker-3B scores 94.3 on AIME26, 80.2 Pass@1 on LiveCodeBench v6, and 96.1% on unseen LeetCode contests. The gains appear to come primarily from post-training on top of
-
Why AI inference prioritises low latency over raw compute
By
–
Why does this matter for AI inference specifically? Training = throughput problem. Inference = latency problem. When a user talks to an AI assistant, tokens have to return fast. Latency, memory access, bandwidth, and interconnect all matter, not just raw compute. In large AI
-
User asks if ChatGPT interface is normal version
By
–
Is this the "normal" version of ChatGPT? The interface looks different from mine.
-
AI false facts caught by anti-hallucination prompt library
By
–
This one prompt has caught more wrong "facts" before I shipped them than any tool I pay for. I keep 1,000+ prompts like it in a free library here: http://
godofprompt.ai/prompt-library
?utm_source=twitter&utm_medium=organic&utm_campaign=antihallucination
… What's something an AI made up that you almost believed? -
Hidden honesty mode to force AIs to state facts
By
–
AI makes up things and says them with a straight face. Fictitious sources, invented numbers, no warning. There is a hidden honesty mode that you can activate with a single copy-paste. It forces ChatGPT, Claude, and Gemini to indicate what is a fact, what is
-
Nemotron 3 Ultra outperforms GPT 5.5 in intelligence
By
–
Don't sleep on Nemotron 3 Ultra Sometimes surprises me that it is more intelligently capable than GPT 5.5
-
Anthropic Talks with Trump Admin Fail, Fable 5 Export Controls Stay
By
–
Update on Fable5/Anthropic: Anthropic flew its top security people to DC. The export controls are still there. Via Wired Anthropic and the Trump administration wrapped up talks on Monday with no resolution – the export controls on Claude Fable 5 are still in place. No end in